praxagent / Work index

Research and systems that hold up in use

Programs, open artifacts, and software built around inspectable evidence, controlled evaluation, and practical operation.

01 / Research programs

Model internals and evaluation

Open fitted lenses, J-space geometry, controlled readout audits, null transports, and revision-pinned receipts.

Feature activation mapping, label-validity tests, balanced contrasts, and reproducible analysis on open models.

A filled remote research mentorship in AI, biology, bioinformatics, technical writing, and public portfolio development.

02 / Open harness

Docs

A security-first self-hosted agentic harness, with an integrated orchestrator, collaboration shell, sandbox, plugins, and keyless egress.

GitHub ↗

Safety-first multi-agent harness with dozens of tools, sophisticated memory management, and more.

Agent-agnostic shell with chat, kanban, files, terminal, and browser screencasts.

Dockerized coding agents, Chromium/CDP, desktop, and a framework-independent Python client.

Self-contained plugin pack for presentations, music, radio, and other runtime extensions.

Isolated credential-injecting egress so a compromised agent has no real keys to steal.

03 / Focused software

An iOS reading system for saved web content, summaries, audio, podcasts, knowledge graphs, and private API-key use.

Interactive strange attractors, fractal morphing, gravitational simulations, and chemical reactions for iOS.